rust/src/librustc/metadata/encoder.rs

1857 lines
63 KiB
Rust
Raw Normal View History

// Copyright 2012 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
2011-06-27 16:03:01 -07:00
// Metadata encoding
use metadata::common::*;
use metadata::cstore;
use metadata::decoder;
use metadata::tyencode;
use middle::ty::{node_id_to_type, lookup_item_type};
use middle::astencode;
use middle::ty;
use middle::typeck;
use middle;
use std::hashmap::{HashMap, HashSet};
use std::io;
use std::str;
use std::vec;
use extra::flate;
use extra::serialize::Encodable;
use extra;
use syntax::abi::AbiSet;
2012-09-04 11:54:36 -07:00
use syntax::ast::*;
use syntax::ast;
2012-09-04 11:54:36 -07:00
use syntax::ast_map;
use syntax::ast_util::*;
2012-09-04 11:54:36 -07:00
use syntax::attr;
use syntax::attr::AttrMetaMethods;
2012-09-04 11:54:36 -07:00
use syntax::diagnostic::span_handler;
use syntax::parse::token::special_idents;
use syntax::ast_util;
use syntax::visit::Visitor;
use syntax::visit;
2013-05-14 17:27:27 -07:00
use syntax::parse::token;
use syntax;
use writer = extra::ebml::writer;
use std::cast;
// used by astencode:
type abbrev_map = @mut HashMap<ty::t, tyencode::ty_abbrev>;
pub type encode_inlined_item<'self> = &'self fn(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
path: &[ast_map::path_elt],
ii: ast::inlined_item);
pub struct EncodeParams<'self> {
diag: @mut span_handler,
tcx: ty::ctxt,
2012-08-29 13:26:26 -07:00
reexports2: middle::resolve::ExportMap2,
item_symbols: &'self HashMap<ast::NodeId, ~str>,
discrim_symbols: &'self HashMap<ast::NodeId, @str>,
non_inlineable_statics: &'self HashSet<ast::NodeId>,
link_meta: &'self LinkMeta,
cstore: @mut cstore::CStore,
encode_inlined_item: encode_inlined_item<'self>,
reachable: @mut HashSet<ast::NodeId>,
}
struct Stats {
inline_bytes: uint,
attr_bytes: uint,
dep_bytes: uint,
lang_item_bytes: uint,
2013-03-06 22:06:53 -08:00
link_args_bytes: uint,
impl_bytes: uint,
misc_bytes: uint,
item_bytes: uint,
index_bytes: uint,
zero_bytes: uint,
total_bytes: uint,
n_inlines: uint
}
2012-08-27 16:53:54 -07:00
pub struct EncodeContext<'self> {
diag: @mut span_handler,
tcx: ty::ctxt,
stats: @mut Stats,
2012-08-29 13:26:26 -07:00
reexports2: middle::resolve::ExportMap2,
item_symbols: &'self HashMap<ast::NodeId, ~str>,
discrim_symbols: &'self HashMap<ast::NodeId, @str>,
non_inlineable_statics: &'self HashSet<ast::NodeId>,
link_meta: &'self LinkMeta,
cstore: &'self cstore::CStore,
encode_inlined_item: encode_inlined_item<'self>,
type_abbrevs: abbrev_map,
reachable: @mut HashSet<ast::NodeId>,
}
pub fn reachable(ecx: &EncodeContext, id: NodeId) -> bool {
2013-03-22 22:26:41 -04:00
ecx.reachable.contains(&id)
}
fn encode_name(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
2013-09-02 02:50:59 +02:00
name: Ident) {
ebml_w.wr_tagged_str(tag_paths_data_name, ecx.tcx.sess.str_of(name));
}
fn encode_impl_type_basename(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
2013-09-02 02:50:59 +02:00
name: Ident) {
ebml_w.wr_tagged_str(tag_item_impl_type_basename,
ecx.tcx.sess.str_of(name));
}
pub fn encode_def_id(ebml_w: &mut writer::Encoder, id: DefId) {
ebml_w.wr_tagged_str(tag_def_id, def_to_str(id));
2012-02-06 07:13:14 -08:00
}
fn encode_region_param(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
it: @ast::item) {
let opt_rp = ecx.tcx.region_paramd_items.find(&it.id);
for rp in opt_rp.iter() {
ebml_w.start_tag(tag_region_param);
rp.encode(ebml_w);
ebml_w.end_tag();
}
}
2013-07-02 12:47:32 -07:00
#[deriving(Clone)]
struct entry<T> {
val: T,
pos: uint
}
2013-06-04 15:14:56 -07:00
fn add_to_index(ebml_w: &mut writer::Encoder,
2013-09-02 02:50:59 +02:00
path: &[Ident],
index: &mut ~[entry<~str>],
2013-09-02 02:50:59 +02:00
name: Ident) {
let mut full_path = ~[];
full_path.push_all(path);
full_path.push(name);
index.push(
entry {
val: ast_util::path_name_i(full_path),
pos: ebml_w.writer.tell()
});
}
fn encode_trait_ref(ebml_w: &mut writer::Encoder,
ecx: &EncodeContext,
trait_ref: &ty::TraitRef,
tag: uint) {
let ty_str_ctxt = @tyencode::ctxt {
diag: ecx.diag,
ds: def_to_str,
tcx: ecx.tcx,
abbrevs: tyencode::ac_use_abbrevs(ecx.type_abbrevs)
};
ebml_w.start_tag(tag);
tyencode::enc_trait_ref(ebml_w.writer, ty_str_ctxt, trait_ref);
ebml_w.end_tag();
}
fn encode_impl_vtables(ebml_w: &mut writer::Encoder,
ecx: &EncodeContext,
vtables: &typeck::impl_res) {
ebml_w.start_tag(tag_item_impl_vtables);
astencode::encode_vtable_res(ecx, ebml_w, vtables.trait_vtables);
astencode::encode_vtable_param_res(ecx, ebml_w, vtables.self_vtables);
ebml_w.end_tag();
}
// Item info table encoding
fn encode_family(ebml_w: &mut writer::Encoder, c: char) {
ebml_w.start_tag(tag_items_data_item_family);
ebml_w.writer.write(&[c as u8]);
ebml_w.end_tag();
}
pub fn def_to_str(did: DefId) -> ~str {
2013-09-27 22:38:08 -07:00
format!("{}:{}", did.crate, did.node)
}
fn encode_ty_type_param_defs(ebml_w: &mut writer::Encoder,
ecx: &EncodeContext,
params: @~[ty::TypeParameterDef],
tag: uint) {
let ty_str_ctxt = @tyencode::ctxt {
diag: ecx.diag,
ds: def_to_str,
tcx: ecx.tcx,
abbrevs: tyencode::ac_use_abbrevs(ecx.type_abbrevs)
};
for param in params.iter() {
ebml_w.start_tag(tag);
tyencode::enc_type_param_def(ebml_w.writer, ty_str_ctxt, param);
ebml_w.end_tag();
}
}
fn encode_bounds_and_type(ebml_w: &mut writer::Encoder,
ecx: &EncodeContext,
tpt: &ty::ty_param_bounds_and_ty) {
encode_ty_type_param_defs(ebml_w, ecx, tpt.generics.type_param_defs,
tag_items_data_item_ty_param_bounds);
encode_type(ecx, ebml_w, tpt.ty);
}
fn encode_variant_id(ebml_w: &mut writer::Encoder, vid: DefId) {
ebml_w.start_tag(tag_items_data_item_variant);
let s = def_to_str(vid);
ebml_w.writer.write(s.as_bytes());
ebml_w.end_tag();
}
pub fn write_type(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
typ: ty::t) {
let ty_str_ctxt = @tyencode::ctxt {
diag: ecx.diag,
ds: def_to_str,
tcx: ecx.tcx,
abbrevs: tyencode::ac_use_abbrevs(ecx.type_abbrevs)
};
tyencode::enc_ty(ebml_w.writer, ty_str_ctxt, typ);
}
pub fn write_vstore(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
vstore: ty::vstore) {
let ty_str_ctxt = @tyencode::ctxt {
diag: ecx.diag,
ds: def_to_str,
tcx: ecx.tcx,
abbrevs: tyencode::ac_use_abbrevs(ecx.type_abbrevs)
};
2012-10-05 16:55:42 -07:00
tyencode::enc_vstore(ebml_w.writer, ty_str_ctxt, vstore);
}
fn encode_type(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
typ: ty::t) {
ebml_w.start_tag(tag_items_data_item_type);
write_type(ecx, ebml_w, typ);
ebml_w.end_tag();
}
fn encode_transformed_self_ty(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
opt_typ: Option<ty::t>) {
for &typ in opt_typ.iter() {
ebml_w.start_tag(tag_item_method_transformed_self_ty);
write_type(ecx, ebml_w, typ);
ebml_w.end_tag();
}
}
fn encode_method_fty(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
typ: &ty::BareFnTy) {
ebml_w.start_tag(tag_item_method_fty);
let ty_str_ctxt = @tyencode::ctxt {
diag: ecx.diag,
ds: def_to_str,
tcx: ecx.tcx,
abbrevs: tyencode::ac_use_abbrevs(ecx.type_abbrevs)
};
tyencode::enc_bare_fn_ty(ebml_w.writer, ty_str_ctxt, typ);
ebml_w.end_tag();
}
fn encode_symbol(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
id: NodeId) {
ebml_w.start_tag(tag_items_data_item_symbol);
match ecx.item_symbols.find(&id) {
2013-03-22 22:26:41 -04:00
Some(x) => {
debug!("encode_symbol(id={:?}, str={})", id, *x);
ebml_w.writer.write(x.as_bytes());
}
None => {
ecx.diag.handler().bug(
2013-09-27 22:38:08 -07:00
format!("encode_symbol: id not found {}", id));
}
}
ebml_w.end_tag();
}
fn encode_disr_val(_: &EncodeContext,
ebml_w: &mut writer::Encoder,
disr_val: ty::Disr) {
ebml_w.start_tag(tag_disr_val);
let s = disr_val.to_str();
ebml_w.writer.write(s.as_bytes());
ebml_w.end_tag();
}
fn encode_parent_item(ebml_w: &mut writer::Encoder, id: DefId) {
ebml_w.start_tag(tag_items_data_parent_item);
let s = def_to_str(id);
ebml_w.writer.write(s.as_bytes());
ebml_w.end_tag();
}
fn encode_struct_fields(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
def: @struct_def) {
for f in def.fields.iter() {
match f.node.kind {
named_field(ident, vis) => {
ebml_w.start_tag(tag_item_field);
encode_struct_field_family(ebml_w, vis);
encode_name(ecx, ebml_w, ident);
encode_def_id(ebml_w, local_def(f.node.id));
ebml_w.end_tag();
}
unnamed_field => {
ebml_w.start_tag(tag_item_unnamed_field);
encode_def_id(ebml_w, local_def(f.node.id));
ebml_w.end_tag();
}
}
}
}
fn encode_enum_variant_info(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
id: NodeId,
variants: &[variant],
path: &[ast_map::path_elt],
2013-08-06 19:35:09 -07:00
index: @mut ~[entry<i64>],
generics: &ast::Generics) {
debug!("encode_enum_variant_info(id={:?})", id);
let mut disr_val = 0;
let mut i = 0;
2013-01-13 11:05:40 -08:00
let vi = ty::enum_variants(ecx.tcx,
ast::DefId { crate: LOCAL_CRATE, node: id });
for variant in variants.iter() {
let def_id = local_def(variant.node.id);
2013-08-06 19:35:09 -07:00
index.push(entry {val: variant.node.id as i64,
pos: ebml_w.writer.tell()});
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
match variant.node.kind {
ast::tuple_variant_kind(_) => encode_family(ebml_w, 'v'),
ast::struct_variant_kind(_) => encode_family(ebml_w, 'V')
}
2012-07-18 16:18:02 -07:00
encode_name(ecx, ebml_w, variant.node.name);
encode_parent_item(ebml_w, local_def(id));
encode_visibility(ebml_w, variant.node.vis);
encode_attributes(ebml_w, variant.node.attrs);
match variant.node.kind {
ast::tuple_variant_kind(ref args)
if args.len() > 0 && generics.ty_params.len() == 0 => {
encode_symbol(ecx, ebml_w, variant.node.id);
}
ast::tuple_variant_kind(_) => {},
ast::struct_variant_kind(def) => {
let idx = encode_info_for_struct(ecx, ebml_w, path,
def.fields, index);
encode_struct_fields(ecx, ebml_w, def);
let bkts = create_index(idx);
encode_index(ebml_w, bkts, write_i64);
}
}
if vi[i].disr_val != disr_val {
encode_disr_val(ecx, ebml_w, vi[i].disr_val);
disr_val = vi[i].disr_val;
}
encode_bounds_and_type(ebml_w, ecx,
&lookup_item_type(ecx.tcx, def_id));
encode_path(ecx, ebml_w, path,
ast_map::path_name(variant.node.name));
ebml_w.end_tag();
disr_val += 1;
i += 1;
}
}
fn encode_path(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
path: &[ast_map::path_elt],
name: ast_map::path_elt) {
fn encode_path_elt(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
2012-07-18 16:18:02 -07:00
elt: ast_map::path_elt) {
match elt {
ast_map::path_mod(n) => {
ebml_w.wr_tagged_str(tag_path_elt_mod, ecx.tcx.sess.str_of(n));
}
ast_map::path_name(n) => {
ebml_w.wr_tagged_str(tag_path_elt_name, ecx.tcx.sess.str_of(n));
}
ast_map::path_pretty_name(n, extra) => {
ebml_w.start_tag(tag_path_elt_pretty_name);
ebml_w.wr_tagged_str(tag_path_elt_pretty_name_ident,
ecx.tcx.sess.str_of(n));
ebml_w.wr_tagged_u64(tag_path_elt_pretty_name_extra, extra);
ebml_w.end_tag();
}
}
}
ebml_w.start_tag(tag_path);
ebml_w.wr_tagged_u32(tag_path_len, (path.len() + 1) as u32);
for pe in path.iter() {
encode_path_elt(ecx, ebml_w, *pe);
}
encode_path_elt(ecx, ebml_w, name);
ebml_w.end_tag();
}
fn encode_reexported_static_method(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
exp: &middle::resolve::Export2,
method_def_id: DefId,
2013-09-02 02:50:59 +02:00
method_ident: Ident) {
debug!("(encode reexported static method) {}::{}",
exp.name, ecx.tcx.sess.str_of(method_ident));
ebml_w.start_tag(tag_items_data_item_reexport);
ebml_w.start_tag(tag_items_data_item_reexport_def_id);
ebml_w.wr_str(def_to_str(method_def_id));
ebml_w.end_tag();
ebml_w.start_tag(tag_items_data_item_reexport_name);
2013-09-27 22:38:08 -07:00
ebml_w.wr_str(format!("{}::{}", exp.name, ecx.tcx.sess.str_of(method_ident)));
ebml_w.end_tag();
ebml_w.end_tag();
}
fn encode_reexported_static_base_methods(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
exp: &middle::resolve::Export2)
-> bool {
match ecx.tcx.inherent_impls.find(&exp.def_id) {
Some(implementations) => {
for &base_impl in implementations.iter() {
for &m in base_impl.methods.iter() {
if m.explicit_self == ast::sty_static {
encode_reexported_static_method(ecx, ebml_w, exp,
m.def_id, m.ident);
}
}
}
true
}
None => { false }
}
}
fn encode_reexported_static_trait_methods(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
exp: &middle::resolve::Export2)
-> bool {
match ecx.tcx.trait_methods_cache.find(&exp.def_id) {
Some(methods) => {
for &m in methods.iter() {
if m.explicit_self == ast::sty_static {
encode_reexported_static_method(ecx, ebml_w, exp,
m.def_id, m.ident);
}
}
true
}
None => { false }
}
}
fn encode_reexported_static_methods(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
mod_path: &[ast_map::path_elt],
exp: &middle::resolve::Export2) {
match ecx.tcx.items.find(&exp.def_id.node) {
Some(&ast_map::node_item(item, path)) => {
let original_name = ecx.tcx.sess.str_of(item.ident);
//
// We don't need to reexport static methods on items
// declared in the same module as our `pub use ...` since
2013-06-01 04:24:31 -07:00
// that's done when we encode the item itself.
//
// The only exception is when the reexport *changes* the
// name e.g. `pub use Foo = self::Bar` -- we have
// encoded metadata for static methods relative to Bar,
// but not yet for Foo.
//
if mod_path != *path || exp.name != original_name {
if !encode_reexported_static_base_methods(ecx, ebml_w, exp) {
if encode_reexported_static_trait_methods(ecx, ebml_w, exp) {
debug!("(encode reexported static methods) {} \
2013-09-27 22:38:08 -07:00
[trait]",
original_name);
}
}
else {
debug!("(encode reexported static methods) {} [base]",
2013-09-27 22:38:08 -07:00
original_name);
}
}
}
_ => {}
}
}
/// Iterates through "auxiliary node IDs", which are node IDs that describe
/// top-level items that are sub-items of the given item. Specifically:
///
/// * For enums, iterates through the node IDs of the variants.
///
/// * For newtype structs, iterates through the node ID of the constructor.
fn each_auxiliary_node_id(item: @item, callback: &fn(NodeId) -> bool)
-> bool {
2013-09-25 17:56:54 -07:00
let mut continue_ = true;
match item.node {
item_enum(ref enum_def, _) => {
for variant in enum_def.variants.iter() {
2013-09-25 17:56:54 -07:00
continue_ = callback(variant.node.id);
if !continue_ {
break
}
}
}
item_struct(struct_def, _) => {
// If this is a newtype struct, return the constructor.
match struct_def.ctor_id {
Some(ctor_id) if struct_def.fields.len() > 0 &&
struct_def.fields[0].node.kind ==
ast::unnamed_field => {
2013-09-25 17:56:54 -07:00
continue_ = callback(ctor_id);
}
_ => {}
}
}
_ => {}
}
2013-09-25 17:56:54 -07:00
continue_
}
fn encode_reexports(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
id: NodeId,
path: &[ast_map::path_elt]) {
debug!("(encoding info for module) encoding reexports for {}", id);
match ecx.reexports2.find(&id) {
Some(ref exports) => {
debug!("(encoding info for module) found reexports for {}", id);
for exp in exports.iter() {
debug!("(encoding info for module) reexport '{}' ({}/{}) for \
2013-09-27 22:38:08 -07:00
{}",
exp.name,
exp.def_id.crate,
exp.def_id.node,
id);
ebml_w.start_tag(tag_items_data_item_reexport);
ebml_w.start_tag(tag_items_data_item_reexport_def_id);
ebml_w.wr_str(def_to_str(exp.def_id));
ebml_w.end_tag();
ebml_w.start_tag(tag_items_data_item_reexport_name);
ebml_w.wr_str(exp.name);
ebml_w.end_tag();
ebml_w.end_tag();
encode_reexported_static_methods(ecx, ebml_w, path, exp);
}
}
None => {
debug!("(encoding info for module) found no reexports for {}",
id);
}
}
}
fn encode_info_for_mod(ecx: &EncodeContext,
2013-05-04 14:25:41 -07:00
ebml_w: &mut writer::Encoder,
md: &_mod,
id: NodeId,
path: &[ast_map::path_elt],
2013-09-02 02:50:59 +02:00
name: Ident,
vis: visibility) {
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, local_def(id));
encode_family(ebml_w, 'm');
2012-07-18 16:18:02 -07:00
encode_name(ecx, ebml_w, name);
debug!("(encoding info for module) encoding info for module ID {}", id);
// Encode info about all the module children.
for item in md.items.iter() {
ebml_w.start_tag(tag_mod_child);
ebml_w.wr_str(def_to_str(local_def(item.id)));
ebml_w.end_tag();
do each_auxiliary_node_id(*item) |auxiliary_node_id| {
ebml_w.start_tag(tag_mod_child);
ebml_w.wr_str(def_to_str(local_def(auxiliary_node_id)));
ebml_w.end_tag();
true
};
match item.node {
item_impl(*) => {
let (ident, did) = (item.ident, item.id);
debug!("(encoding info for module) ... encoding impl {} \
2013-09-27 22:38:08 -07:00
({:?}/{:?})",
ecx.tcx.sess.str_of(ident),
2012-07-18 16:18:02 -07:00
did,
2013-05-14 17:27:27 -07:00
ast_map::node_id_to_str(ecx.tcx.items, did, token::get_ident_interner()));
ebml_w.start_tag(tag_mod_impl);
ebml_w.wr_str(def_to_str(local_def(did)));
ebml_w.end_tag();
}
_ => {}
}
}
encode_path(ecx, ebml_w, path, ast_map::path_mod(name));
2013-10-07 13:01:47 -07:00
encode_visibility(ebml_w, vis);
// Encode the reexports of this module, if this module is public.
if vis == public {
debug!("(encoding info for module) encoding reexports for {}", id);
encode_reexports(ecx, ebml_w, id, path);
}
ebml_w.end_tag();
}
2013-05-04 14:25:41 -07:00
fn encode_struct_field_family(ebml_w: &mut writer::Encoder,
visibility: visibility) {
2012-08-06 12:34:08 -07:00
encode_family(ebml_w, match visibility {
2012-08-03 19:59:04 -07:00
public => 'g',
private => 'j',
inherited => 'N'
});
}
fn encode_visibility(ebml_w: &mut writer::Encoder, visibility: visibility) {
ebml_w.start_tag(tag_items_data_item_visibility);
let ch = match visibility {
public => 'y',
private => 'n',
inherited => 'i',
};
ebml_w.wr_str(str::from_char(ch));
ebml_w.end_tag();
}
fn encode_explicit_self(ebml_w: &mut writer::Encoder, explicit_self: ast::explicit_self_) {
ebml_w.start_tag(tag_item_trait_method_explicit_self);
// Encode the base self type.
match explicit_self {
sty_static => {
ebml_w.writer.write(&[ 's' as u8 ]);
}
sty_value(m) => {
ebml_w.writer.write(&[ 'v' as u8 ]);
encode_mutability(ebml_w, m);
}
sty_region(_, m) => {
// FIXME(#4846) encode custom lifetime
ebml_w.writer.write(&[ '&' as u8 ]);
encode_mutability(ebml_w, m);
}
sty_box(m) => {
ebml_w.writer.write(&[ '@' as u8 ]);
encode_mutability(ebml_w, m);
}
sty_uniq(m) => {
ebml_w.writer.write(&[ '~' as u8 ]);
encode_mutability(ebml_w, m);
}
}
ebml_w.end_tag();
fn encode_mutability(ebml_w: &writer::Encoder,
m: ast::Mutability) {
match m {
MutImmutable => ebml_w.writer.write(&[ 'i' as u8 ]),
MutMutable => ebml_w.writer.write(&[ 'm' as u8 ]),
}
}
}
fn encode_method_sort(ebml_w: &mut writer::Encoder, sort: char) {
ebml_w.start_tag(tag_item_trait_method_sort);
ebml_w.writer.write(&[ sort as u8 ]);
ebml_w.end_tag();
}
fn encode_provided_source(ebml_w: &mut writer::Encoder,
source_opt: Option<DefId>) {
for source in source_opt.iter() {
ebml_w.start_tag(tag_item_method_provided_source);
let s = def_to_str(*source);
ebml_w.writer.write(s.as_bytes());
ebml_w.end_tag();
}
}
/* Returns an index of items in this class */
fn encode_info_for_struct(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
path: &[ast_map::path_elt],
fields: &[@struct_field],
2013-08-06 19:35:09 -07:00
global_index: @mut ~[entry<i64>])
-> ~[entry<i64>] {
/* Each class has its own index, since different classes
may have fields with the same name */
2013-07-02 12:47:32 -07:00
let mut index = ~[];
let tcx = ecx.tcx;
/* We encode both private and public fields -- need to include
private fields to get the offsets right */
for field in fields.iter() {
let (nm, vis) = match field.node.kind {
named_field(nm, vis) => (nm, vis),
unnamed_field => (special_idents::unnamed_field, inherited)
};
let id = field.node.id;
2013-08-06 19:35:09 -07:00
index.push(entry {val: id as i64, pos: ebml_w.writer.tell()});
global_index.push(entry {val: id as i64, pos: ebml_w.writer.tell()});
ebml_w.start_tag(tag_items_data_item);
debug!("encode_info_for_struct: doing {} {}",
tcx.sess.str_of(nm), id);
encode_struct_field_family(ebml_w, vis);
encode_name(ecx, ebml_w, nm);
encode_path(ecx, ebml_w, path, ast_map::path_name(nm));
encode_type(ecx, ebml_w, node_id_to_type(tcx, id));
encode_def_id(ebml_w, local_def(id));
ebml_w.end_tag();
}
2013-07-02 12:47:32 -07:00
index
}
fn encode_info_for_struct_ctor(ecx: &EncodeContext,
2013-05-04 14:25:41 -07:00
ebml_w: &mut writer::Encoder,
path: &[ast_map::path_elt],
2013-09-02 02:50:59 +02:00
name: ast::Ident,
ctor_id: NodeId,
index: @mut ~[entry<i64>],
struct_id: NodeId) {
2013-08-06 19:35:09 -07:00
index.push(entry { val: ctor_id as i64, pos: ebml_w.writer.tell() });
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, local_def(ctor_id));
encode_family(ebml_w, 'f');
encode_name(ecx, ebml_w, name);
encode_type(ecx, ebml_w, node_id_to_type(ecx.tcx, ctor_id));
encode_path(ecx, ebml_w, path, ast_map::path_name(name));
encode_parent_item(ebml_w, local_def(struct_id));
if ecx.item_symbols.contains_key(&ctor_id) {
encode_symbol(ecx, ebml_w, ctor_id);
}
ebml_w.end_tag();
}
fn encode_method_ty_fields(ecx: &EncodeContext,
2013-05-04 14:25:41 -07:00
ebml_w: &mut writer::Encoder,
method_ty: &ty::Method) {
encode_def_id(ebml_w, method_ty.def_id);
encode_name(ecx, ebml_w, method_ty.ident);
encode_ty_type_param_defs(ebml_w, ecx,
method_ty.generics.type_param_defs,
tag_item_method_tps);
encode_transformed_self_ty(ecx, ebml_w, method_ty.transformed_self_ty);
encode_method_fty(ecx, ebml_w, &method_ty.fty);
encode_visibility(ebml_w, method_ty.vis);
encode_explicit_self(ebml_w, method_ty.explicit_self);
let purity = method_ty.fty.purity;
match method_ty.explicit_self {
ast::sty_static => {
encode_family(ebml_w, purity_static_method_family(purity));
}
_ => encode_family(ebml_w, purity_fn_family(purity))
}
encode_provided_source(ebml_w, method_ty.provided_source);
}
fn encode_info_for_method(ecx: &EncodeContext,
2013-05-04 14:25:41 -07:00
ebml_w: &mut writer::Encoder,
m: &ty::Method,
impl_path: &[ast_map::path_elt],
is_default_impl: bool,
parent_id: NodeId,
ast_method_opt: Option<@method>) {
debug!("encode_info_for_method: {:?} {}", m.def_id,
ecx.tcx.sess.str_of(m.ident));
ebml_w.start_tag(tag_items_data_item);
encode_method_ty_fields(ecx, ebml_w, m);
encode_parent_item(ebml_w, local_def(parent_id));
// The type for methods gets encoded twice, which is unfortunate.
let tpt = lookup_item_type(ecx.tcx, m.def_id);
encode_bounds_and_type(ebml_w, ecx, &tpt);
encode_path(ecx, ebml_w, impl_path, ast_map::path_name(m.ident));
for ast_method in ast_method_opt.iter() {
let num_params = tpt.generics.type_param_defs.len();
if num_params > 0u || is_default_impl
|| should_inline(ast_method.attrs) {
(ecx.encode_inlined_item)(
ecx, ebml_w, impl_path,
ii_method(local_def(parent_id), false, *ast_method));
} else {
encode_symbol(ecx, ebml_w, m.def_id.node);
}
}
ebml_w.end_tag();
}
fn purity_fn_family(p: purity) -> char {
2012-08-06 12:34:08 -07:00
match p {
2012-08-03 19:59:04 -07:00
unsafe_fn => 'u',
impure_fn => 'f',
extern_fn => 'e'
}
}
fn purity_static_method_family(p: purity) -> char {
match p {
unsafe_fn => 'U',
impure_fn => 'F',
_ => fail!("extern fn can't be static")
2012-02-10 13:39:15 -08:00
}
}
fn should_inline(attrs: &[Attribute]) -> bool {
use syntax::attr::*;
match find_inline_attr(attrs) {
InlineNone | InlineNever => false,
InlineHint | InlineAlways => true
}
}
// Encodes the inherent implementations of a structure, enumeration, or trait.
fn encode_inherent_implementations(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
def_id: DefId) {
match ecx.tcx.inherent_impls.find(&def_id) {
None => {}
Some(&implementations) => {
for implementation in implementations.iter() {
ebml_w.start_tag(tag_items_data_item_inherent_impl);
encode_def_id(ebml_w, implementation.did);
ebml_w.end_tag();
}
}
}
}
// Encodes the implementations of a trait defined in this crate.
fn encode_extension_implementations(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
trait_def_id: DefId) {
match ecx.tcx.trait_impls.find(&trait_def_id) {
None => {}
Some(&implementations) => {
for implementation in implementations.iter() {
ebml_w.start_tag(tag_items_data_item_extension_impl);
encode_def_id(ebml_w, implementation.did);
ebml_w.end_tag();
}
}
}
}
fn encode_info_for_item(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
item: @item,
2013-08-06 19:35:09 -07:00
index: @mut ~[entry<i64>],
path: &[ast_map::path_elt],
vis: ast::visibility) {
let tcx = ecx.tcx;
fn add_to_index_(item: @item, ebml_w: &writer::Encoder,
2013-08-06 19:35:09 -07:00
index: @mut ~[entry<i64>]) {
index.push(entry { val: item.id as i64, pos: ebml_w.writer.tell() });
}
let add_to_index: &fn() = || add_to_index_(item, ebml_w, index);
debug!("encoding info for item at {}",
2012-11-12 18:24:56 -08:00
ecx.tcx.sess.codemap.span_to_str(item.span));
let def_id = local_def(item.id);
match item.node {
2013-06-21 18:46:34 -07:00
item_static(_, m, _) => {
add_to_index();
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
if m == ast::MutMutable {
2013-06-21 18:46:34 -07:00
encode_family(ebml_w, 'b');
} else {
encode_family(ebml_w, 'c');
}
encode_type(ecx, ebml_w, node_id_to_type(tcx, item.id));
2011-07-27 14:19:39 +02:00
encode_symbol(ecx, ebml_w, item.id);
encode_name(ecx, ebml_w, item.ident);
let elt = ast_map::path_pretty_name(item.ident, item.id as u64);
encode_path(ecx, ebml_w, path, elt);
if !ecx.non_inlineable_statics.contains(&item.id) {
(ecx.encode_inlined_item)(ecx, ebml_w, path, ii_item(item));
}
encode_visibility(ebml_w, vis);
ebml_w.end_tag();
2011-07-27 14:19:39 +02:00
}
item_fn(_, purity, _, ref generics, _) => {
add_to_index();
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
encode_family(ebml_w, purity_fn_family(purity));
let tps_len = generics.ty_params.len();
encode_bounds_and_type(ebml_w, ecx, &lookup_item_type(tcx, def_id));
encode_name(ecx, ebml_w, item.ident);
encode_path(ecx, ebml_w, path, ast_map::path_name(item.ident));
encode_attributes(ebml_w, item.attrs);
if tps_len > 0u || should_inline(item.attrs) {
(ecx.encode_inlined_item)(ecx, ebml_w, path, ii_item(item));
} else {
encode_symbol(ecx, ebml_w, item.id);
}
encode_visibility(ebml_w, vis);
ebml_w.end_tag();
2011-07-27 14:19:39 +02:00
}
item_mod(ref m) => {
add_to_index();
encode_info_for_mod(ecx,
ebml_w,
m,
item.id,
path,
item.ident,
item.vis);
2011-07-27 14:19:39 +02:00
}
item_foreign_mod(ref fm) => {
add_to_index();
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
encode_family(ebml_w, 'n');
2012-07-18 16:18:02 -07:00
encode_name(ecx, ebml_w, item.ident);
encode_path(ecx, ebml_w, path, ast_map::path_name(item.ident));
// Encode all the items in this module.
for foreign_item in fm.items.iter() {
ebml_w.start_tag(tag_mod_child);
ebml_w.wr_str(def_to_str(local_def(foreign_item.id)));
ebml_w.end_tag();
}
encode_visibility(ebml_w, vis);
ebml_w.end_tag();
2011-07-27 14:19:39 +02:00
}
item_ty(*) => {
add_to_index();
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
encode_family(ebml_w, 'y');
encode_bounds_and_type(ebml_w, ecx, &lookup_item_type(tcx, def_id));
2012-07-18 16:18:02 -07:00
encode_name(ecx, ebml_w, item.ident);
encode_path(ecx, ebml_w, path, ast_map::path_name(item.ident));
encode_region_param(ecx, ebml_w, item);
encode_visibility(ebml_w, vis);
ebml_w.end_tag();
2011-07-27 14:19:39 +02:00
}
item_enum(ref enum_definition, ref generics) => {
add_to_index();
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
encode_family(ebml_w, 't');
encode_bounds_and_type(ebml_w, ecx, &lookup_item_type(tcx, def_id));
encode_name(ecx, ebml_w, item.ident);
for v in (*enum_definition).variants.iter() {
encode_variant_id(ebml_w, local_def(v.node.id));
}
(ecx.encode_inlined_item)(ecx, ebml_w, path, ii_item(item));
encode_path(ecx, ebml_w, path, ast_map::path_name(item.ident));
encode_region_param(ecx, ebml_w, item);
// Encode inherent implementations for this enumeration.
encode_inherent_implementations(ecx, ebml_w, def_id);
encode_visibility(ebml_w, vis);
ebml_w.end_tag();
2012-12-04 21:13:02 -08:00
encode_enum_variant_info(ecx,
ebml_w,
item.id,
(*enum_definition).variants,
2012-12-04 21:13:02 -08:00
path,
index,
generics);
2011-07-27 14:19:39 +02:00
}
item_struct(struct_def, _) => {
/* First, encode the fields
These come first because we need to write them to make
the index, and the index needs to be in the item for the
class itself */
let idx = encode_info_for_struct(ecx, ebml_w, path,
struct_def.fields, index);
/* Index the class*/
add_to_index();
/* Now, make an item for the class itself */
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
encode_family(ebml_w, 'S');
encode_bounds_and_type(ebml_w, ecx, &lookup_item_type(tcx, def_id));
2012-07-18 16:18:02 -07:00
encode_name(ecx, ebml_w, item.ident);
2013-05-06 23:35:27 +09:00
encode_attributes(ebml_w, item.attrs);
encode_path(ecx, ebml_w, path, ast_map::path_name(item.ident));
encode_region_param(ecx, ebml_w, item);
encode_visibility(ebml_w, vis);
/* Encode def_ids for each field and method
for methods, write all the stuff get_trait_method
needs to know*/
encode_struct_fields(ecx, ebml_w, struct_def);
(ecx.encode_inlined_item)(ecx, ebml_w, path, ii_item(item));
// Encode inherent implementations for this structure.
encode_inherent_implementations(ecx, ebml_w, def_id);
/* Each class has its own index -- encode it */
let bkts = create_index(idx);
2013-08-06 19:35:09 -07:00
encode_index(ebml_w, bkts, write_i64);
ebml_w.end_tag();
// If this is a tuple- or enum-like struct, encode the type of the
// constructor.
if struct_def.fields.len() > 0 &&
struct_def.fields[0].node.kind == ast::unnamed_field {
let ctor_id = match struct_def.ctor_id {
Some(ctor_id) => ctor_id,
None => ecx.tcx.sess.bug("struct def didn't have ctor id"),
};
encode_info_for_struct_ctor(ecx,
ebml_w,
path,
item.ident,
ctor_id,
index,
def_id.node);
}
}
item_impl(_, ref opt_trait, ref ty, ref ast_methods) => {
// We need to encode information about the default methods we
// have inherited, so we drive this based on the impl structure.
let imp = tcx.impls.get(&def_id);
add_to_index();
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
encode_family(ebml_w, 'i');
encode_region_param(ecx, ebml_w, item);
encode_bounds_and_type(ebml_w, ecx, &lookup_item_type(tcx, def_id));
2012-07-18 16:18:02 -07:00
encode_name(ecx, ebml_w, item.ident);
encode_attributes(ebml_w, item.attrs);
match ty.node {
ast::ty_path(ref path, ref bounds, _) if path.segments
.len() == 1 => {
assert!(bounds.is_none());
encode_impl_type_basename(ecx, ebml_w,
ast_util::path_to_ident(path));
}
_ => {}
}
for method in imp.methods.iter() {
ebml_w.start_tag(tag_item_impl_method);
let s = def_to_str(method.def_id);
ebml_w.writer.write(s.as_bytes());
ebml_w.end_tag();
}
for ast_trait_ref in opt_trait.iter() {
let trait_ref = ty::node_id_to_trait_ref(
tcx, ast_trait_ref.ref_id);
encode_trait_ref(ebml_w, ecx, trait_ref, tag_item_trait_ref);
let impl_vtables = ty::lookup_impl_vtables(tcx, def_id);
encode_impl_vtables(ebml_w, ecx, &impl_vtables);
}
let elt = ast_map::impl_pretty_name(opt_trait, ty, item.ident);
encode_path(ecx, ebml_w, path, elt);
ebml_w.end_tag();
// >:-<
let mut impl_path = vec::append(~[], path);
impl_path.push(elt);
// Iterate down the methods, emitting them. We rely on the
// assumption that all of the actually implemented methods
// appear first in the impl structure, in the same order they do
// in the ast. This is a little sketchy.
let num_implemented_methods = ast_methods.len();
for (i, m) in imp.methods.iter().enumerate() {
let ast_method = if i < num_implemented_methods {
Some(ast_methods[i])
} else { None };
2013-08-06 19:35:09 -07:00
index.push(entry {val: m.def_id.node as i64,
pos: ebml_w.writer.tell()});
encode_info_for_method(ecx,
ebml_w,
*m,
impl_path,
false,
item.id,
ast_method)
}
}
item_trait(_, ref super_traits, ref ms) => {
add_to_index();
ebml_w.start_tag(tag_items_data_item);
encode_def_id(ebml_w, def_id);
encode_family(ebml_w, 'I');
encode_region_param(ecx, ebml_w, item);
let trait_def = ty::lookup_trait_def(tcx, def_id);
encode_ty_type_param_defs(ebml_w, ecx,
trait_def.generics.type_param_defs,
tag_items_data_item_ty_param_bounds);
encode_trait_ref(ebml_w, ecx, trait_def.trait_ref, tag_item_trait_ref);
2012-07-18 16:18:02 -07:00
encode_name(ecx, ebml_w, item.ident);
encode_attributes(ebml_w, item.attrs);
for &method_def_id in ty::trait_method_def_ids(tcx, def_id).iter() {
ebml_w.start_tag(tag_item_trait_method);
encode_def_id(ebml_w, method_def_id);
ebml_w.end_tag();
ebml_w.start_tag(tag_mod_child);
ebml_w.wr_str(def_to_str(method_def_id));
ebml_w.end_tag();
}
encode_path(ecx, ebml_w, path, ast_map::path_name(item.ident));
// FIXME(#8559): This should use the tcx's supertrait cache instead of
// reading the AST's list, because the former has already filtered out
// the builtin-kinds-as-supertraits. See corresponding fixme in decoder.
for ast_trait_ref in super_traits.iter() {
let trait_ref = ty::node_id_to_trait_ref(ecx.tcx, ast_trait_ref.ref_id);
encode_trait_ref(ebml_w, ecx, trait_ref, tag_item_super_trait_ref);
}
// Encode the implementations of this trait.
encode_extension_implementations(ecx, ebml_w, def_id);
ebml_w.end_tag();
// Now output the method info for each method.
let r = ty::trait_method_def_ids(tcx, def_id);
for (i, &method_def_id) in r.iter().enumerate() {
assert_eq!(method_def_id.crate, ast::LOCAL_CRATE);
let method_ty = ty::method(tcx, method_def_id);
2013-08-06 19:35:09 -07:00
index.push(entry {val: method_def_id.node as i64,
pos: ebml_w.writer.tell()});
ebml_w.start_tag(tag_items_data_item);
encode_method_ty_fields(ecx, ebml_w, method_ty);
encode_parent_item(ebml_w, def_id);
let mut trait_path = vec::append(~[], path);
trait_path.push(ast_map::path_name(item.ident));
encode_path(ecx, ebml_w, trait_path, ast_map::path_name(method_ty.ident));
match method_ty.explicit_self {
sty_static => {
encode_family(ebml_w,
purity_static_method_family(
method_ty.fty.purity));
let tpt = ty::lookup_item_type(tcx, method_def_id);
encode_bounds_and_type(ebml_w, ecx, &tpt);
}
_ => {
encode_family(ebml_w,
purity_fn_family(
method_ty.fty.purity));
}
}
match ms[i] {
required(_) => {
encode_method_sort(ebml_w, 'r');
}
provided(m) => {
// If this is a static method, we've already encoded
// this.
if method_ty.explicit_self != sty_static {
// XXX: I feel like there is something funny going on.
let tpt = ty::lookup_item_type(tcx, method_def_id);
encode_bounds_and_type(ebml_w, ecx, &tpt);
}
encode_method_sort(ebml_w, 'p');
(ecx.encode_inlined_item)(
ecx, ebml_w, path,
ii_method(def_id, true, m));
}
}
ebml_w.end_tag();
}
// Encode inherent implementations for this trait.
encode_inherent_implementations(ecx, ebml_w, def_id);
}
item_mac(*) => fail!("item macros unimplemented")
}
}
fn encode_info_for_foreign_item(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
nitem: @foreign_item,
2013-08-06 19:35:09 -07:00
index: @mut ~[entry<i64>],
2013-07-02 12:47:32 -07:00
path: &ast_map::path,
abi: AbiSet) {
2013-08-06 19:35:09 -07:00
index.push(entry { val: nitem.id as i64, pos: ebml_w.writer.tell() });
ebml_w.start_tag(tag_items_data_item);
match nitem.node {
foreign_item_fn(*) => {
2011-07-27 14:19:39 +02:00
encode_def_id(ebml_w, local_def(nitem.id));
encode_family(ebml_w, purity_fn_family(impure_fn));
encode_bounds_and_type(ebml_w, ecx,
&lookup_item_type(ecx.tcx,local_def(nitem.id)));
encode_name(ecx, ebml_w, nitem.ident);
if abi.is_intrinsic() {
2013-07-02 12:47:32 -07:00
(ecx.encode_inlined_item)(ecx, ebml_w, *path, ii_foreign(nitem));
} else {
encode_symbol(ecx, ebml_w, nitem.id);
}
2013-07-02 12:47:32 -07:00
encode_path(ecx, ebml_w, *path, ast_map::path_name(nitem.ident));
2011-07-27 14:19:39 +02:00
}
foreign_item_static(_, mutbl) => {
encode_def_id(ebml_w, local_def(nitem.id));
if mutbl {
encode_family(ebml_w, 'b');
} else {
encode_family(ebml_w, 'c');
}
encode_type(ecx, ebml_w, node_id_to_type(ecx.tcx, nitem.id));
encode_symbol(ecx, ebml_w, nitem.id);
encode_name(ecx, ebml_w, nitem.ident);
2013-07-02 12:47:32 -07:00
encode_path(ecx, ebml_w, *path, ast_map::path_name(nitem.ident));
}
}
ebml_w.end_tag();
}
fn my_visit_expr(_e:@Expr) { }
fn my_visit_item(i:@item, items: ast_map::map, ebml_w:&writer::Encoder,
ecx_ptr:*int, index: @mut ~[entry<i64>]) {
match items.get_copy(&i.id) {
ast_map::node_item(_, pt) => {
let mut ebml_w = ebml_w.clone();
// See above
let ecx : &EncodeContext = unsafe { cast::transmute(ecx_ptr) };
Extract privacy checking from name resolution This commit is the culmination of my recent effort to refine Rust's notion of privacy and visibility among crates. The major goals of this commit were to remove privacy checking from resolve for the sake of sane error messages, and to attempt a much more rigid and well-tested implementation of visibility throughout rust. The implemented rules for name visibility are: 1. Everything pub from the root namespace is visible to anyone 2. You may access any private item of your ancestors. "Accessing a private item" depends on what the item is, so for a function this means that you can call it, but for a module it means that you can look inside of it. Once you look inside a private module, any accessed item must be "pub from the root" where the new root is the private module that you looked into. These rules required some more analysis results to get propagated from trans to privacy in the form of a few hash tables. I added a new test in which my goal was to showcase all of the privacy nuances of the language, and I hope to place any new bugs into this file to prevent regressions. Overall, I was unable to completely remove the notion of privacy from resolve. One use of privacy is for dealing with glob imports. Essentially a glob import can only import *public* items from the destination, and because this must be done at namespace resolution time, resolve must maintain the notion of "what items are public in a module". There are some sad approximations of privacy, but I unfortunately can't see clear methods to extract them outside. The other use case of privacy in resolve now is one that must stick around regardless of glob imports. When dealing with privacy, checking a private path needs to know "what the last private thing was" when looking at a path. Resolve is the only compiler pass which knows the answer to this question, so it maintains the answer on a per-path resolution basis (works similarly to the def_map generated). Closes #8215
2013-10-05 14:37:39 -07:00
encode_info_for_item(ecx, &mut ebml_w, i, index, *pt, i.vis);
}
_ => fail!("bad item")
}
}
fn my_visit_foreign_item(ni:@foreign_item, items: ast_map::map, ebml_w:&writer::Encoder,
ecx_ptr:*int, index: @mut ~[entry<i64>]) {
match items.get_copy(&ni.id) {
ast_map::node_foreign_item(_, abi, _, pt) => {
debug!("writing foreign item {}::{}",
ast_map::path_to_str(
*pt,
token::get_ident_interner()),
token::ident_to_str(&ni.ident));
let mut ebml_w = ebml_w.clone();
// See above
let ecx : &EncodeContext = unsafe { cast::transmute(ecx_ptr) };
encode_info_for_foreign_item(ecx,
&mut ebml_w,
ni,
index,
pt,
abi);
}
// case for separate item and foreign-item tables
_ => fail!("bad foreign item")
}
}
struct EncodeVisitor {
ebml_w_for_visit_item: writer::Encoder,
ebml_w_for_visit_foreign_item: writer::Encoder,
ecx_ptr:*int,
items: ast_map::map,
index: @mut ~[entry<i64>],
}
impl visit::Visitor<()> for EncodeVisitor {
fn visit_expr(&mut self, ex:@Expr, _:()) {
visit::walk_expr(self, ex, ());
my_visit_expr(ex);
}
fn visit_item(&mut self, i:@item, _:()) {
visit::walk_item(self, i, ());
my_visit_item(i,
self.items,
&self.ebml_w_for_visit_item,
self.ecx_ptr,
self.index);
}
fn visit_foreign_item(&mut self, ni:@foreign_item, _:()) {
visit::walk_foreign_item(self, ni, ());
my_visit_foreign_item(ni,
self.items,
&self.ebml_w_for_visit_foreign_item,
self.ecx_ptr,
self.index);
}
}
fn encode_info_for_items(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
crate: &Crate)
2013-08-06 19:35:09 -07:00
-> ~[entry<i64>] {
let index = @mut ~[];
ebml_w.start_tag(tag_items_data);
2013-08-06 19:35:09 -07:00
index.push(entry { val: CRATE_NODE_ID as i64, pos: ebml_w.writer.tell() });
encode_info_for_mod(ecx,
ebml_w,
&crate.module,
CRATE_NODE_ID,
[],
syntax::parse::token::special_idents::invalid,
public);
let items = ecx.tcx.items;
// See comment in `encode_side_tables_for_ii` in astencode
let ecx_ptr : *int = unsafe { cast::transmute(ecx) };
let mut visitor = EncodeVisitor {
index: index,
items: items,
ecx_ptr: ecx_ptr,
ebml_w_for_visit_item: (*ebml_w).clone(),
ebml_w_for_visit_foreign_item: (*ebml_w).clone(),
};
visit::walk_crate(&mut visitor, crate, ());
ebml_w.end_tag();
2013-07-02 12:47:32 -07:00
return /*bad*/(*index).clone();
}
// Path and definition ID indexing
fn create_index<T:Clone + Hash + IterBytes + 'static>(
index: ~[entry<T>])
-> ~[@~[entry<T>]] {
let mut buckets: ~[@mut ~[entry<T>]] = ~[];
for _ in range(0u, 256u) { buckets.push(@mut ~[]); };
for elt in index.iter() {
let h = elt.val.hash() as uint;
2013-07-02 12:47:32 -07:00
buckets[h % 256].push((*elt).clone());
}
let mut buckets_frozen = ~[];
for bucket in buckets.iter() {
2013-07-02 12:47:32 -07:00
buckets_frozen.push(@/*bad*/(**bucket).clone());
}
2012-08-01 17:30:05 -07:00
return buckets_frozen;
}
fn encode_index<T:'static>(
ebml_w: &mut writer::Encoder,
buckets: ~[@~[entry<T>]],
write_fn: &fn(@io::Writer, &T)) {
let writer = ebml_w.writer;
ebml_w.start_tag(tag_index);
let mut bucket_locs: ~[uint] = ~[];
ebml_w.start_tag(tag_index_buckets);
for bucket in buckets.iter() {
bucket_locs.push(ebml_w.writer.tell());
ebml_w.start_tag(tag_index_buckets_bucket);
for elt in (**bucket).iter() {
ebml_w.start_tag(tag_index_buckets_bucket_elt);
2013-03-28 18:39:09 -07:00
assert!(elt.pos < 0xffff_ffff);
writer.write_be_u32(elt.pos as u32);
write_fn(writer, &elt.val);
ebml_w.end_tag();
}
ebml_w.end_tag();
}
ebml_w.end_tag();
ebml_w.start_tag(tag_index_table);
for pos in bucket_locs.iter() {
2013-03-28 18:39:09 -07:00
assert!(*pos < 0xffff_ffff);
writer.write_be_u32(*pos as u32);
}
ebml_w.end_tag();
ebml_w.end_tag();
}
fn write_str(writer: @io::Writer, s: ~str) {
writer.write_str(s);
}
2013-08-06 19:35:09 -07:00
fn write_i64(writer: @io::Writer, &n: &i64) {
2013-03-28 18:39:09 -07:00
assert!(n < 0x7fff_ffff);
writer.write_be_u32(n as u32);
}
fn encode_meta_item(ebml_w: &mut writer::Encoder, mi: @MetaItem) {
2012-08-06 12:34:08 -07:00
match mi.node {
MetaWord(name) => {
ebml_w.start_tag(tag_meta_item_word);
ebml_w.start_tag(tag_meta_item_name);
ebml_w.writer.write(name.as_bytes());
ebml_w.end_tag();
ebml_w.end_tag();
2011-07-27 14:19:39 +02:00
}
MetaNameValue(name, value) => {
2012-08-06 12:34:08 -07:00
match value.node {
lit_str(value, _) => {
ebml_w.start_tag(tag_meta_item_name_value);
ebml_w.start_tag(tag_meta_item_name);
ebml_w.writer.write(name.as_bytes());
ebml_w.end_tag();
ebml_w.start_tag(tag_meta_item_value);
ebml_w.writer.write(value.as_bytes());
ebml_w.end_tag();
ebml_w.end_tag();
2011-07-27 14:19:39 +02:00
}
2012-08-03 19:59:04 -07:00
_ => {/* FIXME (#623): encode other variants */ }
2011-07-27 14:19:39 +02:00
}
}
MetaList(name, ref items) => {
ebml_w.start_tag(tag_meta_item_list);
ebml_w.start_tag(tag_meta_item_name);
ebml_w.writer.write(name.as_bytes());
ebml_w.end_tag();
for inner_item in items.iter() {
encode_meta_item(ebml_w, *inner_item);
}
ebml_w.end_tag();
2011-07-27 14:19:39 +02:00
}
}
}
fn encode_attributes(ebml_w: &mut writer::Encoder, attrs: &[Attribute]) {
ebml_w.start_tag(tag_attributes);
for attr in attrs.iter() {
ebml_w.start_tag(tag_attribute);
encode_meta_item(ebml_w, attr.node.value);
ebml_w.end_tag();
}
ebml_w.end_tag();
}
// So there's a special crate attribute called 'link' which defines the
// metadata that Rust cares about for linking crates. This attribute requires
2011-06-30 17:03:08 -07:00
// 'name' and 'vers' items, so if the user didn't provide them we will throw
// them in anyway with default values.
fn synthesize_crate_attrs(ecx: &EncodeContext,
crate: &Crate) -> ~[Attribute] {
fn synthesize_link_attr(ecx: &EncodeContext, items: ~[@MetaItem]) ->
Attribute {
2013-03-28 18:39:09 -07:00
assert!(!ecx.link_meta.name.is_empty());
assert!(!ecx.link_meta.vers.is_empty());
2011-07-27 14:19:39 +02:00
let name_item =
attr::mk_name_value_item_str(@"name",
ecx.link_meta.name);
2011-07-27 14:19:39 +02:00
let vers_item =
attr::mk_name_value_item_str(@"vers",
ecx.link_meta.vers);
let mut meta_items = ~[name_item, vers_item];
for &mi in items.iter().filter(|mi| "name" != mi.name() && "vers" != mi.name()) {
meta_items.push(mi);
}
let link_item = attr::mk_list_item(@"link", meta_items);
2012-08-01 17:30:05 -07:00
return attr::mk_attr(link_item);
}
let mut attrs = ~[];
let mut found_link_attr = false;
for attr in crate.attrs.iter() {
attrs.push(
if "link" != attr.name() {
*attr
2011-07-27 14:19:39 +02:00
} else {
match attr.meta_item_list() {
Some(l) => {
found_link_attr = true;;
synthesize_link_attr(ecx, l.to_owned())
2011-07-27 14:19:39 +02:00
}
_ => *attr
}
});
}
if !found_link_attr { attrs.push(synthesize_link_attr(ecx, ~[])); }
2012-08-01 17:30:05 -07:00
return attrs;
}
fn encode_crate_deps(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
cstore: &cstore::CStore) {
fn get_ordered_deps(ecx: &EncodeContext, cstore: &cstore::CStore)
-> ~[decoder::CrateDep] {
type numdep = decoder::CrateDep;
// Pull the cnums and name,vers,hash out of cstore
2013-02-11 08:36:42 -08:00
let mut deps = ~[];
2012-06-30 16:19:07 -07:00
do cstore::iter_crate_data(cstore) |key, val| {
let dep = decoder::CrateDep {cnum: key,
name: ecx.tcx.sess.ident_of(val.name),
vers: decoder::get_crate_vers(val.data),
2012-06-10 00:49:59 -07:00
hash: decoder::get_crate_hash(val.data)};
deps.push(dep);
};
// Sort by cnum
extra::sort::quick_sort(deps, |kv1, kv2| kv1.cnum <= kv2.cnum);
// Sanity-check the crate numbers
let mut expected_cnum = 1;
for n in deps.iter() {
assert_eq!(n.cnum, expected_cnum);
expected_cnum += 1;
}
deps
}
// We're just going to write a list of crate 'name-hash-version's, with
// the assumption that they are numbered 1 to n.
// FIXME (#2166): This is not nearly enough to support correct versioning
// but is enough to get transitive crate dependencies working.
ebml_w.start_tag(tag_crate_deps);
let r = get_ordered_deps(ecx, cstore);
for dep in r.iter() {
encode_crate_dep(ecx, ebml_w, *dep);
}
ebml_w.end_tag();
}
fn encode_lang_items(ecx: &EncodeContext, ebml_w: &mut writer::Encoder) {
ebml_w.start_tag(tag_lang_items);
for (i, def_id) in ecx.tcx.lang_items.items() {
for id in def_id.iter() {
if id.crate == LOCAL_CRATE {
ebml_w.start_tag(tag_lang_items_item);
ebml_w.start_tag(tag_lang_items_item_id);
ebml_w.writer.write_be_u32(i as u32);
ebml_w.end_tag(); // tag_lang_items_item_id
ebml_w.start_tag(tag_lang_items_item_node_id);
ebml_w.writer.write_be_u32(id.node as u32);
ebml_w.end_tag(); // tag_lang_items_item_node_id
ebml_w.end_tag(); // tag_lang_items_item
}
}
}
ebml_w.end_tag(); // tag_lang_items
}
fn encode_link_args(ecx: &EncodeContext, ebml_w: &mut writer::Encoder) {
2013-03-06 22:06:53 -08:00
ebml_w.start_tag(tag_link_args);
let link_args = cstore::get_used_link_args(ecx.cstore);
for link_arg in link_args.iter() {
2013-03-06 22:06:53 -08:00
ebml_w.start_tag(tag_link_args_arg);
ebml_w.writer.write_str(link_arg.to_str());
ebml_w.end_tag();
}
ebml_w.end_tag();
}
struct ImplVisitor<'self> {
ecx: &'self EncodeContext<'self>,
ebml_w: &'self mut writer::Encoder,
}
impl<'self> Visitor<()> for ImplVisitor<'self> {
fn visit_item(&mut self, item: @item, _: ()) {
match item.node {
item_impl(_, Some(ref trait_ref), _, _) => {
let def_map = self.ecx.tcx.def_map;
let trait_def = def_map.get_copy(&trait_ref.ref_id);
let def_id = ast_util::def_id_of_def(trait_def);
// Load eagerly if this is an implementation of the Drop trait
// or if the trait is not defined in this crate.
Extract privacy checking from name resolution This commit is the culmination of my recent effort to refine Rust's notion of privacy and visibility among crates. The major goals of this commit were to remove privacy checking from resolve for the sake of sane error messages, and to attempt a much more rigid and well-tested implementation of visibility throughout rust. The implemented rules for name visibility are: 1. Everything pub from the root namespace is visible to anyone 2. You may access any private item of your ancestors. "Accessing a private item" depends on what the item is, so for a function this means that you can call it, but for a module it means that you can look inside of it. Once you look inside a private module, any accessed item must be "pub from the root" where the new root is the private module that you looked into. These rules required some more analysis results to get propagated from trans to privacy in the form of a few hash tables. I added a new test in which my goal was to showcase all of the privacy nuances of the language, and I hope to place any new bugs into this file to prevent regressions. Overall, I was unable to completely remove the notion of privacy from resolve. One use of privacy is for dealing with glob imports. Essentially a glob import can only import *public* items from the destination, and because this must be done at namespace resolution time, resolve must maintain the notion of "what items are public in a module". There are some sad approximations of privacy, but I unfortunately can't see clear methods to extract them outside. The other use case of privacy in resolve now is one that must stick around regardless of glob imports. When dealing with privacy, checking a private path needs to know "what the last private thing was" when looking at a path. Resolve is the only compiler pass which knows the answer to this question, so it maintains the answer on a per-path resolution basis (works similarly to the def_map generated). Closes #8215
2013-10-05 14:37:39 -07:00
if Some(def_id) == self.ecx.tcx.lang_items.drop_trait() ||
def_id.crate != LOCAL_CRATE {
self.ebml_w.start_tag(tag_impls_impl);
encode_def_id(self.ebml_w, local_def(item.id));
self.ebml_w.end_tag();
}
}
_ => {}
}
visit::walk_item(self, item, ());
}
}
/// Encodes implementations that are eagerly loaded.
///
/// None of this is necessary in theory; we can load all implementations
/// lazily. However, in two cases the optimizations to lazily load
/// implementations are not yet implemented. These two cases, which require us
/// to load implementations eagerly, are:
///
/// * Destructors (implementations of the Drop trait).
///
/// * Implementations of traits not defined in this crate.
fn encode_impls(ecx: &EncodeContext,
crate: &Crate,
ebml_w: &mut writer::Encoder) {
ebml_w.start_tag(tag_impls);
{
let mut visitor = ImplVisitor {
ecx: ecx,
ebml_w: ebml_w,
};
visit::walk_crate(&mut visitor, crate, ());
}
ebml_w.end_tag();
}
fn encode_misc_info(ecx: &EncodeContext,
crate: &Crate,
ebml_w: &mut writer::Encoder) {
ebml_w.start_tag(tag_misc_info);
ebml_w.start_tag(tag_misc_info_crate_items);
for &item in crate.module.items.iter() {
ebml_w.start_tag(tag_mod_child);
ebml_w.wr_str(def_to_str(local_def(item.id)));
ebml_w.end_tag();
do each_auxiliary_node_id(item) |auxiliary_node_id| {
ebml_w.start_tag(tag_mod_child);
ebml_w.wr_str(def_to_str(local_def(auxiliary_node_id)));
ebml_w.end_tag();
true
};
}
// Encode reexports for the root module.
encode_reexports(ecx, ebml_w, 0, []);
ebml_w.end_tag();
ebml_w.end_tag();
}
fn encode_crate_dep(ecx: &EncodeContext,
ebml_w: &mut writer::Encoder,
dep: decoder::CrateDep) {
ebml_w.start_tag(tag_crate_dep);
ebml_w.start_tag(tag_crate_dep_name);
let s = ecx.tcx.sess.str_of(dep.name);
ebml_w.writer.write(s.as_bytes());
ebml_w.end_tag();
ebml_w.start_tag(tag_crate_dep_vers);
ebml_w.writer.write(dep.vers.as_bytes());
ebml_w.end_tag();
ebml_w.start_tag(tag_crate_dep_hash);
ebml_w.writer.write(dep.hash.as_bytes());
ebml_w.end_tag();
ebml_w.end_tag();
}
fn encode_hash(ebml_w: &mut writer::Encoder, hash: &str) {
ebml_w.start_tag(tag_crate_hash);
ebml_w.writer.write(hash.as_bytes());
ebml_w.end_tag();
2011-12-11 23:23:38 +08:00
}
// NB: Increment this as you change the metadata encoding version.
pub static metadata_encoding_version : &'static [u8] =
&[0x72, //'r' as u8,
0x75, //'u' as u8,
0x73, //'s' as u8,
0x74, //'t' as u8,
0, 0, 0, 1 ];
pub fn encode_metadata(parms: EncodeParams, crate: &Crate) -> ~[u8] {
let wr = @io::BytesWriter::new();
2013-04-12 01:15:30 -04:00
let stats = Stats {
inline_bytes: 0,
attr_bytes: 0,
dep_bytes: 0,
lang_item_bytes: 0,
2013-03-06 22:06:53 -08:00
link_args_bytes: 0,
impl_bytes: 0,
misc_bytes: 0,
item_bytes: 0,
index_bytes: 0,
zero_bytes: 0,
total_bytes: 0,
n_inlines: 0
};
let EncodeParams {
item_symbols,
diag,
tcx,
reexports2,
discrim_symbols,
cstore,
encode_inlined_item,
link_meta,
reachable,
non_inlineable_statics,
_
} = parms;
let type_abbrevs = @mut HashMap::new();
let stats = @mut stats;
let ecx = EncodeContext {
2013-03-22 22:26:41 -04:00
diag: diag,
tcx: tcx,
stats: stats,
2013-03-22 22:26:41 -04:00
reexports2: reexports2,
item_symbols: item_symbols,
discrim_symbols: discrim_symbols,
non_inlineable_statics: non_inlineable_statics,
2013-03-22 22:26:41 -04:00
link_meta: link_meta,
cstore: cstore,
encode_inlined_item: encode_inlined_item,
type_abbrevs: type_abbrevs,
reachable: reachable,
};
let mut ebml_w = writer::Encoder(wr as @io::Writer);
encode_hash(&mut ebml_w, ecx.link_meta.extras_hash);
2011-12-11 23:23:38 +08:00
let mut i = *wr.pos;
let crate_attrs = synthesize_crate_attrs(&ecx, crate);
encode_attributes(&mut ebml_w, crate_attrs);
ecx.stats.attr_bytes = *wr.pos - i;
i = *wr.pos;
encode_crate_deps(&ecx, &mut ebml_w, ecx.cstore);
ecx.stats.dep_bytes = *wr.pos - i;
// Encode the language items.
i = *wr.pos;
encode_lang_items(&ecx, &mut ebml_w);
ecx.stats.lang_item_bytes = *wr.pos - i;
2013-03-06 22:06:53 -08:00
// Encode the link args.
i = *wr.pos;
encode_link_args(&ecx, &mut ebml_w);
ecx.stats.link_args_bytes = *wr.pos - i;
2013-03-06 22:06:53 -08:00
// Encode the def IDs of impls, for coherence checking.
i = *wr.pos;
encode_impls(&ecx, crate, &mut ebml_w);
ecx.stats.impl_bytes = *wr.pos - i;
// Encode miscellaneous info.
i = *wr.pos;
encode_misc_info(&ecx, crate, &mut ebml_w);
ecx.stats.misc_bytes = *wr.pos - i;
// Encode and index the items.
ebml_w.start_tag(tag_items);
i = *wr.pos;
let items_index = encode_info_for_items(&ecx, &mut ebml_w, crate);
ecx.stats.item_bytes = *wr.pos - i;
2012-08-27 16:53:54 -07:00
i = *wr.pos;
let items_buckets = create_index(items_index);
2013-08-06 19:35:09 -07:00
encode_index(&mut ebml_w, items_buckets, write_i64);
ecx.stats.index_bytes = *wr.pos - i;
ebml_w.end_tag();
ecx.stats.total_bytes = *wr.pos;
2012-08-27 16:53:54 -07:00
2013-03-22 22:26:41 -04:00
if (tcx.sess.meta_stats()) {
for e in wr.bytes.iter() {
2013-03-07 18:37:14 -05:00
if *e == 0 {
ecx.stats.zero_bytes += 1;
2012-08-27 16:53:54 -07:00
}
}
io::println("metadata stats:");
println!(" inline bytes: {}", ecx.stats.inline_bytes);
println!(" attribute bytes: {}", ecx.stats.attr_bytes);
println!(" dep bytes: {}", ecx.stats.dep_bytes);
println!(" lang item bytes: {}", ecx.stats.lang_item_bytes);
println!(" link args bytes: {}", ecx.stats.link_args_bytes);
println!(" impl bytes: {}", ecx.stats.impl_bytes);
println!(" misc bytes: {}", ecx.stats.misc_bytes);
println!(" item bytes: {}", ecx.stats.item_bytes);
println!(" index bytes: {}", ecx.stats.index_bytes);
println!(" zero bytes: {}", ecx.stats.zero_bytes);
println!(" total bytes: {}", ecx.stats.total_bytes);
2012-08-27 16:53:54 -07:00
}
// Pad this, since something (LLVM, presumably) is cutting off the
// remaining % 4 bytes.
wr.write(&[0u8, 0u8, 0u8, 0u8]);
let writer_bytes: &mut ~[u8] = wr.bytes;
metadata_encoding_version.to_owned() +
flate::deflate_bytes(*writer_bytes)
}
2011-06-27 19:16:16 -07:00
// Get the encoded string for a type
pub fn encoded_ty(tcx: ty::ctxt, t: ty::t) -> ~str {
let cx = @tyencode::ctxt {
diag: tcx.diag,
ds: def_to_str,
tcx: tcx,
abbrevs: tyencode::ac_no_abbrevs};
do io::with_str_writer |wr| {
tyencode::enc_ty(wr, cx, t);
}
}